bug: Wrong executed amounts in competition data

[fhenneke]

Problem

It seems that the data on the buy and sell amounts of orders added in #1975 is not accurate.
This is only implemented on staging if I understand correctly.

For example, the transaction with hash 0x2548006a1bf02717d4e0a20bd6f0db7e8762d146161135f2d22f857eb726740d gives executed amounts

but the sell amount on etherscan

and our order book endpoint

are different.

Impact

Having the correct data in the competition endpoint is required for EBBO tests.

Expected behaviour

I expect that the "sellAmount" and "buyAmount" in the competition endpoint to be the amounts send and received by a user in that settlement.

[fleupold]

It looks like the amount shown there is the sellAmount + solver fee (which doesn't matter here because it's a "market" order where the fee is specified by the order itself: log

This seems to be because the driver assumes the fee is correctly solver specified if it's present on the response of the solver engine:

services/crates/driver/src/infra/solver/dto/solution.rs

Lines 41 to 49 in 03ed19b

	competition::solution::trade::Fulfillment::new(
	order,
	fulfillment.executed_amount.into(),
	match fulfillment.fee {
	Some(fee) => competition::solution::trade::Fee::Dynamic(
	competition::order::SellAmount(fee),
	),
	None => competition::solution::trade::Fee::Static,
	},

We can either check there is the order has a solver defined fee, or fix it in the solver engines to not set this value for order types that don't expect it. The more defensive thing imo would be to change it in the driver.

[fleupold]

Actually I think my assessment from yesterday was incorrect, we do already reject settlements from solvers that set a fee on orders they aren't supposed to. The discrepancy between the amount you see and executedSellAmount in the API is the difference between the subsidized and unsubsidized fee amount.

The executed amount is adding the "solver fee" (this name is used with many different semantics throughout the codebase, but here it's supposed to mean the fee that the solver should use when evaluating the objective value which factors our the 0.002 ETH subsidy we give on market orders), but it should be adding the "user fee" instead.

#2151 should address the issue.

I assume it's ok to report the fully executed buy and sell amounts (after all fees) for the EBBO test.

[fhenneke]

I assume it's ok to report the fully executed buy and sell amounts (after all fees) for the EBBO test.

Yes, this is exactly what we need: the amount users send us (i.e. before deducting any fees in the sell token) and the amount users receive (i.e. after deducting any fees in the buy token).